A new approach to robot guidance by understanding the Indication-Posts is proposed. The approach is based on the conviction that once a human being is in an unfamiliar environment he can arrive at a destination by following the IPs encountered.
Only a monocular image of the scene taken by the robot and local information are required in this approach. To realize this concept, several IPs depicting simple information were demonstrated. To understand the meaning depicted by the IP, image processing and pattern recognition techniques were integrated to resolve the problems of IP finding, robot location determination, IP identification and understanding. Experimental results in a simulated environment verified that the approach is feasible and reliable.